How to change group number on SG cluster device

[Mike Smith_33]

Newbie sysadmin that I am, I have added a vg group into my ServiceGuard 11.16 on HPUX 11.11 cluster. This now prevents me from running make_tape_recovery.

I need the steps to safely change the group number. The device that I managed to duplicate was my cluster lock device - vglck.

It is a two node cluster, here is the output
from

prod node

# ll /dev/*/group
crw-r----- 1 root sys 64 0x000000 Jun 29 01:34 /dev/vg00/group
crw-rw-rw- 1 root sys 64 0x010000 Sep 14 05:14 /dev/vg01/group
crw-rw-rw- 1 root sys 64 0x020000 Sep 14 11:25 /dev/vg02/group
crw-rw-rw- 1 root sys 64 0x030000 Sep 14 11:33 /dev/vg03/group
crw-rw-rw- 1 root sys 64 0x040000 Sep 14 11:34 /dev/vg04/group
crw------- 1 root sys 64 0x050000 Oct 14 17:25 /dev/vg05/group
crw-rw-rw- 1 root sys 64 0x050000 Sep 20 14:11 /dev/vglck/group


Here is output from dev node

# ll /dev/*/group
crw-r----- 1 root sys 64 0x000000 Sep 28 12:47 /dev/vg00/group
crw-rw-rw- 1 root sys 64 0x010000 Sep 22 12:19 /dev/vg01/group
crw-rw-rw- 1 root sys 64 0x060000 Sep 22 11:45 /dev/vg01_dev/grp
crw-rw-rw- 1 root sys 64 0x020000 Sep 22 12:26 /dev/vg02/group
crw-rw-rw- 1 root sys 64 0x070000 Sep 22 11:30 /dev/vg02_dev/grp
crw-rw-rw- 1 root sys 64 0x030000 Sep 22 12:27 /dev/vg03/group
crw-rw-rw- 1 root sys 64 0x080000 Sep 22 12:05 /dev/vg03_dev/grp
crw------- 1 root sys 64 0x040000 Nov 2 15:20 /dev/vg04/group
crw------- 1 root sys 64 0x0a0000 Nov 2 14:42 /dev/vg04_dev/grp
crw------- 1 root sys 64 0x050000 Oct 20 16:29 /dev/vg05/group
crw------- 1 root sys 64 0x090000 Oct 20 19:40 /dev/vg05_dev/grp
crw-rw-rw- 1 root sys 64 0x050000 Sep 20 14:16 /dev/vglck/group


I would appreciate some help on this one.

[Luk Vandenbussche]

Hi

you have to change the kernel parameter
maxvgs on both nodes.

Then you can create a group file with a higher number then a (hexadecimal) or 10 (decimal)

[Mike Smith_33]

Actually I did that I while back when I added the extra vg devices.

[Kent Ostby]

Mike --

Were you actually able to add the new vg into lvmtab?

I would think it would have puked with you trying to add a second group with the same group ID .

If all you have done is add the group to /dev/vg05/*, then it can removed, renamed, and then the vg re-imported on both nodes (or if you change an unshared /dev/vg05, it will only need to be changed on the one node).

[Luk Vandenbussche]

Can you send the cluster config file?

If vglck contains your cluster lock then i think you can solve it by simply removing vg05 on both nodes

vgexport vg05

[Mike Smith_33]

I agree that I am suprised the whole thing did not croak. One thing I need to add, this is not something I did yesterday, it has been over a month. I have run make_tape_recovery sucessfully since then. It doesn't hang every time, only some times. HP told me I needed to change the minor number for one of them, they apparently assumed I knew how to do this.

Here is my dilemma, vg05 has data and is actively used but the cluster is not in production yet. I would like to save the data so it would seem that changing vglck would be easier to do except since it is the cluster device, I am not sure what the steps are.

[Duncan Edmonstone · Accepted solution]

Actually changing vg05 would be easier as presumably vglck contains the cluster lock disk yes?

You don't need to lose the data - vgexport doesn't delete any data from disks, just removes configuration information from /etc/lvmtab and device files from /dev - just do the following:

On first node:

1) Stop the package that uses vg05 using cmhaltpkg

2) Create a map file for vg05:

vgexport -p -s -m /tmp/vg05.map

3) Get rid of vg05 from lvmtab:

vgexport vg05

4) Re-create vg05:

mkdir /dev/vg05
mknod /dev/vg05/group c 64 0x100000
vgimport -s -m /tmp/vg05.map /dev/vg05

5) Copy /tmp/vg05.map to 2nd node.

On second node, repeat steps 3 and 4.

HTH

Duncan

[Duncan Edmonstone]

oops - step 2 should of course be:

vgexport -p -s -m /tmp/vg05.map /dev/vg05

HTH

Duncan

[Mike Smith_33]

In step four, shouldnt the group number be more like 0x0b0000?

[Geoff Wild]

The group minor can be whatever is available - just as long as it doesn't exist already....

Another tip - I would stick to a vg naming standard - don't name them vg02_dev - stick with numbers - a whole lot less confusing...

That way your group minor numbers will align (in hex) nicely with the vgXX

IE:

vg01 group minor 0x010000
.
.
vg10 group minor 0x0a0000

etc...

Rgds...Geoff

[Mike Smith_33]

# vgimport -p -s -m /tmp/vg05.map /dev/vg05
vgimport: Cannot open the control file "/dev/vg05/group":
No such device


I followed the steps as listed.

[Geoff Wild]

ll /dev/vg05/group

If not there, then you need to mknod it

mkdir /dev/vg05
mknod /dev/vg05/group c 64 0x0b0000


Rgds...Geoff

[Mike Smith_33]

Ok, I found the problem. I had only increase maxvgs on the dev node not the prod node. I changed this in the kernel, processed it and rebooted. Everything cool now.

[Geoff Wild]

Glad you got it...now - as far as maintianing environments on your cluster - talk to your HP Support Rep about CCMON - the tool is free - but it costs you a Technical Services day to have it installed (HP has to install it the first time).

http://h71028.www7.hp.com/enterprise/downloads/ccmon-service-brief.pdf

Basically, it monitors differneces in your cluster - everything from kernel parms, vgs, to individual files that you can specify...

Rgds...Geoff